The Association between Social Vulnerability and Geriatric Assessment Impairments among Older Adults with Gastrointestinal Cancers – The CARE Registry

Abstract

Background

Older adults comprise the majority of gastrointestinal (GI) cancer cases. Geriatric assessments (GA) are recommended for older adults with cancer in part to detect aging-related impairments (e.g. frailty), associated with early mortality. Social factors like social vulnerability may also influence aging-related impairments. However, the association between social vulnerability and aging outcomes among older adults with cancer is understudied.

Methods

We included 908 older adults ≥60y recently diagnosed with GI cancer undergoing GA at first pre-chemotherapy visit to UAB oncology clinic. Exposure was Social Vulnerability Index (SVI). Outcomes were frailty (frail vs. robust/pre-frail) and total GA impairments (range: 0–13). We examined the association between SVI and outcomes using Poisson regression with robust variance estimation and generalized estimating equations.

Results

Median age at GA was 69y (IQR: 64, 75), 58.2% were male, 22.6% non-Hispanic Black, 29.1% had colorectal and 28.2% pancreatic cancer, and 70.3% had stage III/IV disease. Adjusting for age, sex, cancer type and stage, each decile increase in SVI was associated with 8% higher prevalence of frailty (PR: 1.08, 95% CI: 1.05, 1.11) and 4% higher average count of total GA impairments (RR: 1.04, 95% CI: 1.02, 1.06). Results were attenuated after further adjustment for race and education.

Conclusions

Greater social vulnerability is associated with higher prevalence of frailty and increasing average GA impairments among older adults with GI cancers prior to systemic treatment. Intervening on social vulnerability may be targets for improving risk of frailty and GA impairments, but associations of race and education should be further evaluated.

PRECIS

Higher social vulnerability at the census tract level is associated with higher prevalence of frailty and risk of additional GA impairments among older adults with GI malignancies prior to systemic treatment. These associations are attenuated with adjustment for individual-level measures of social determinants of health, but multi-level interventions may be a target to prevent these important aging-related outcomes in cancer.

INTRODUCTION

The United States population is aging.¹ Cancer risk increases with older age and older adults represent the majority of cancer cases.² In fact, by 2040, nearly 70% of incident cancer cases will be among older adults and the number of adults ≥85 years old developing cancer is anticipated to more than double in the United States.³ Gastrointestinal (GI) cancers are one of the groups of cancers with highest prevalence among older adults.⁴ The management of older adults with GI malignancies is complex due to the prevalence and risk of aging-related impairments such as frailty, functional decline, comorbidities, and others.⁵ Current guidelines recommend use of geriatric assessment (GA) in the clinic setting for evaluation and management of frailty and other aging-related syndromes in cancer, which can increase risk of chemotherapy toxicity and mortality.^6–10

Development of frailty, functional decline, and other GA impairments is affected by a milieu of upstream components such as genetics and environment.^11,12 Specifically, social determinants of health (SDOH) are environmental factors reflecting the “conditions in the environments where people are born, live, learn, work, play, worship, and age that affect a wide range of health, functioning, and quality-of-life outcomes and risks”. SDOH encompass five domains: 1.) economic stability; 2.) education access and quality; 3.) health care access and quality; 4.) neighborhood and built environment; and 5.) social and community context with an array of area- and person-level variables in each domain.¹³ However, composite measures of SDOH also exist. One such composite measure of SDOH is the social vulnerability index (SVI).^14,15

SVI has demonstrated its ability to predict mortality, treatment receipt, and surgical outcomes in the GI cancer setting.^16,17 However, despite this, the association between SVI and more proximal predictors of mortality like frailty and other GA-identified impairments among older adults with GI cancers is under-studied. Therefore, our objective was to evaluate the association between social vulnerability and frailty among older adults with GI cancers.

MATERIALS AND METHODS

Study Population

We included participants from the Cancer & Aging Resilience Evaluation (CARE) Registry at the University of Alabama at Birmingham (UAB) enrolled from September 2017 to March 2023.¹⁸ For the current study, we included 908 older adults (≥60 years) newly diagnosed with GI cancers who completed a patient-reported pragmatic GA prior to systemic treatment at their first visit to the medical oncology clinic at UAB, an academic medical center in the Southeastern United States (US). We excluded those in the active or post-treatment phase due to inability to distinguish whether any association between SVI and outcomes was driven by treatment type/phase or existed prior to treatment (and then exacerbated by treatment). We chose 60 years as the age cutoff in lieu of the traditional definition of “older adult” as 65 years in recognition of uncertainty around the “right” age cutoff and to allow for meaningful age-related sub-analyses. The CARE GA was conducted via paper survey from September 2017 to December 2020 at which point it was transitioned to electronic capture via implementation of the Web-Enabled Cancer & Aging Resilience Evaluation (WeCARE) across several medical oncology sub-specialties for clinical use among patients across the lifespan. The CARE GA includes self-reported measures of cognition, function, physical performance, nutrition, falls, anxiety and depression, comorbidities, social activities, health-related quality of life (HRQoL), and socioeconomic/demographic characteristics.^19–29 (Supplementary Table 1). The UAB Institutional Review Board approved this study. All procedures were conducted in accordance with ethical standards and principles set forth by the Declaration of Helsinki. The data underlying this article cannot be shared to protect the privacy of study participants. However, summary-level data may be shared upon request to the authors.

Exposure

We measured the primary exposure, social vulnerability, using the 2020 SVI. SVI was initially developed by the Centers for Disease Control and Prevention (CDC) Agency for Toxic Substances and Disease Registry (ATSDR) to assist agencies in identifying communities with greatest support needs in the event of a natural or other disaster.^14,15 However, it has recently gained attention in medical research as a predictor for negative health outcomes.³⁰ SVI is a reflection of the potential for negative effects on communities caused by external stressors to human health and encompasses many variables across themes such as racial and ethnic minority status, socioeconomic status, household characteristics, and housing type/transportation.^14,15,30 SVI is an index measure of SDOH encompassing several SDOH variables. SVI utilizes 16 social factors from the US Census Bureau to rank census tracts according to their relative social vulnerability. SVI ranks all US census tracts on the 16 social factors across domains such as unemployment, racial/ethnic minority status, disability, and others. SVI can further be sub-divided into 4 sub-themes according to the census variables: socioeconomic status, household characteristics, racial/ethnic minority status, and housing type and transportation. Therefore, each census tract has a rank for the census variable, the theme-level SVI, and the overall SVI. Detailed methodology for calculating SVI is located elsewhere including a list of census variables used for ranking theme-level and overall SVI.^14,15 Both overall and theme-level SVI rankings range from 0 to 1 with 0 representing the 0^th percentile and 1 representing the 100^th percentile for lowest and highest social vulnerability, respectively. In the current analysis, we treated SVI as a continuous variable.

Outcomes

Primary Outcome: Frailty

We determined frailty, the primary outcome, via the CARE Frailty Index (CARE FI).³¹ The CARE FI is a measure with demonstrated predictive validity for clinical outcomes such as mortality, treatment toxicity, and functional decline among older adults with GI cancers.³¹ The CARE FI was calculated via a 44-item index based on the principles of deficit accumulation. Detailed methodology for calculating the CARE FI is described elsewhere and in Appendix A in the Supplementary Material.³¹ Briefly, across GA domains, each item queried presence or absence of a health deficit and was scored as 0 (absence of deficit), 0.5 (intermediate response [e.g. sometimes/maybe]), or 1 (presence of deficit). Individual items were totaled and divided by total number of items to obtain CARE FI ranging from 0–1 (0=no deficits, 1=44/44 deficits). Scores were then categorized using standard threshold scoring: robust (<0.2), pre-frail (0.2–0.35), frail (>0.35).^32–34 CARE FI was calculated only for participants with ≥30 items completed. For the current analysis, we dichotomized frailty as frail versus pre-frail/robust.³⁵

Secondary Outcome: Number of GA Impairments

To obtain the secondary outcome of number of GA impairments, we totaled the number of impaired measures across the GA. Total impairments can range from 0 (no measures impaired measure) to 13 (all measures impaired) impairments. Specific measures within each GA domain are detailed in Supplementary Table 1.

Covariates

We included demographics (age at GA completion, race/ethnicity, sex, marital status, living arrangements), socioeconomic variables (education and employment status), and clinical characteristics (cancer type and stage) as covariates. Race/ethnicity, sex, marital status, living arrangements, education, and employment status were self-reported. Age, cancer type, and cancer stage were abstracted from the electronic medical record. We also included rural versus urban residence as a covariate. Rural-urban status was defined by Rural-Urban Commuting Area (RUCA) codes of the participant’s census tract of residence which were classified into rural or urban using Categorization C developed by the University of Washington’s Rural Health Research Center. Detailed methods for this categorization are available elsewhere.³⁶

Statistical Analysis

We completed descriptive statistics for demographics, socioeconomic variables, and clinical variables using frequency and proportions and median (interquartile range) for categorical and continuous variables, respectively. In Supplementary Table 2, bivariate descriptive statistics are also presented by SVI quartiles using chi-square tests and t-tests for categorical and continuous variables, respectively. We used Fisher’s exact test for categorical variables with insufficient expected cell size and the Mann Whitney-U test for continuous variables with non-normal distribution.

We evaluated the association between SVI and frailty using modified Poisson regression with robust variance estimation and generalized estimating equations to account for census tract clustering. Modified Poisson regression also to obtains a measure of risk as opposed to odds from traditional logistic regression. Given the high prevalence of frailty in the sample, the odds ratio likely over-estimates the risk and odds ratios are also frequently mis-interpreted by readers in the medical literature.^37,38 We evaluated the association between SVI and total GA impairments using traditional Poisson regression with generalized estimating equations with an exchangeable correlation matrix. Total GA impairments were not over- or under-dispersed and the model fit well with a QIC of −240.5.³⁹ We adjusted models for age, sex, cancer type, and cancer stage. We conducted a second, third, and fourth set of models further adjusting for individual-level race, education, and rural-urban status, respectively. SVI was parameterized such that the prevalence ratio (PR) or risk ratio (RR) corresponded to a 0.1-point, or 10% increase, in SVI. We also examined highest versus lowest SVI, 1 (highest vulnerability) versus 0 (lowest vulnerability). We chose model covariates a priori and based on bivariate analyses. We also evaluated interaction between SVI and education level. All tests were two-sided with significance set at α=0.05 and were performed using SAS Version 9.4 (SAS Institute, Inc., Cary, NC).

RESULTS

Of 2039 participants in the total sample, 355 were excluded due to missing SVI, 422 were excluded due to cancer of a non-GI organ, and 354 were excluded due to being in active or post-chemotherapy treatment phase leaving a final analytic sample of 908 (Figure 1). Overall, median (interquartile range, [IQR]) SVI of the study sample was 0.58 (0.28, 0.76) and 307 (34.9%) were experiencing frailty. The median age of the study sample was 69 years (IQR: 64, 75), 58.2% were male, 22.6% were Non-Hispanic (NH)-Black, 41.6% had less than or equal to a high school education, 60.5% were married, 62.8% were retired, 61.4% lived with a spouse, and 24.5% lived in a rural area. A majority of the sample had either colorectal (29.1%) or pancreatic (28.2%) cancers and most were advanced stage disease (III: 29.2%, IV: 41.1%) [Table 1]. Differences between those included and excluded from the sample are presented in Supplementary Table 3. A small proportion more participants identifying as NH-Black or Other race-ethnicity groups were included in the sample (NH-Black: 22.6% vs. 21.7%; Other: 4.2% vs. 2.0%; p-value: 0.021). As expected, since only patients with GI cancers were included in this sample, there was a statistically significant difference in the cancer types included versus those excluded. A higher proportion of participants with stage IV disease were excluded (41.1% vs. 55.0%, p-value: <0.001). Median overall SVI was slightly higher for excluded participants (0.514 vs. 0.582, p-value: 0.003) and median SVI for theme 4 (housing type and transportation) was higher among excluded participants (0.373 vs. 0.437, p-value: <0.001). However, having a non-missing value for SVI was an exclusion criterion so some in the excluded category were missing these variables. A higher proportion of frail participants were included (34.9% vs. 27.4%, p-value: <0.001). Finally, while statistically significant, there was no difference in median total GA impairments (2 vs. 2, p-value <0.001).

Figure 1. Flow of Participants to arrive at Final Analytic Sample.

Table 1.

Descriptive Statistics of the Overall Study Sample (n=908)

Variable	Summary Measure
Demographics
Age, median (IQR)	69 (64, 75)
Sex, male, n (%)	528 (58.2)
Race/Ethnicity, n (%)
NH-White	650 (73.3)
NH-Black	200 (22.6)
Other	37 (4.2)
Education, n (%)
<High School	120 (14.0)
High School	236 (27.6)
Associate’s/Bachelor’s	366 (42.8)
Advanced Degree	133 (15.6)
Marital Status, n (%)
Single/Widowed/Divorced	333 (39.5)
Married	510 (60.5)
Employment Status, n (%)
Retired	506 (62.8)
Disabled	109 (13.5)
Working	115 (14.3)
Other	76 (9.4)
Living Arrangement, n (%)
Spouse	473 (61.4)
Alone	176 (22.8)
Other	122 (15.8)
Rural-Urban Status, n (%)
Rural	222 (24.5)
Urban	686 (75.6)
Clinical Characteristics
Cancer Type, n (%)
Colorectal	264 (29.1)
Pancreatic	256 (28.2)
Hepatobiliary	176 (19.4)
Esophageal/Gastric	91 (10.0)
Other	121 (13.3)
Cancer Stage, n (%)
I/II	263 (29.7)
III	259 (29.2)
IV	364 (41.1)
Exposure
Overall SVI, median (IQR)	0.58 (0.28, 0.76)
SVI Theme 1, median (IQR)	0.63 (0.35, 0.81)
SVI Theme 2, median (IQR)	0.69 (0.43, 0.86)
SVI Theme 3, median (IQR)	0.38 (0.21, 0.54)
SVI Theme 4, median (IQR)	0.44 (0.23, 0.70)
Outcomes
Frailty, n (%)	307 (34.9)
Frailty, 3 categories, n (%)
Robust	318 (36.1)
Pre-Frail	255 (29.0)
Frail	307 (34.9)
Total GA Impairments, median (IQR)	2 (1, 4)

Abbreviations: NH-White=Non-Hispanic White; NH-Black=Non-Hispanic Black; SVI=Social Vulnerability Index; GA=Geriatric Assessment

SVI Theme 1: Socioeconomic Status

SVI Theme 2: Household Characteristics

SVI Theme 3: Racial / Ethnic Minority Status

SVI Theme 4: Housing Type / Transportation

SVI obtained by geocoding street addresses to census tracts and matching to the Centers for Disease Control and Prevention SVI database. SVI ranks census tracts on 16 social factors and produces an overall SVI rank from 0 to 1 and individual ranks for each theme also from 0 to 1. 0 indicates lowest social vulnerability and 1 represents highest social vulnerability^14,15

Overall, after adjustment for age, sex, cancer type, and stage, a 10% increase in SVI was associated with a 8% higher prevalence of frailty (PR: 1.08, 95% Confidence Interval [95% CI]): 1.05, 1.11) and a 4% average increase in count of total GA impairments (RR: 1.04, 95% CI: 1.02, 1.06). Examining each SVI theme separately, the socioeconomic status theme demonstrated the greatest magnitude of association for prevalence of frailty (Socioeconomic Status PR [95% CI]: 1.07 [1.03, 1.10); Household Characteristics PR [95% CI]: 1.06 [1.03, 1.10]; Racial/Ethnic Minority Status PR [95% CI]: 1.06 [1.02, 1.10]; Housing Type and Transportation: 1.04 [1.01, 1.08]) and socioeconomic status and racial/ethnic minority status demonstrating the greatest magnitude of association for average increase in count of of total GA impairments (Socioeconomic Status RR [95% CI]: 1.04 [1.02, 1.07); Household Characteristics RR [95% CI]: 1.03 [1.01, 1.06]; Racial/Ethnic Minority Status RR [95% CI]: 1.04 [1.02, 1.07]; Housing Type and Transportation RR [95% CI]: 1.02 [1.00, 1.04]) [Table 2]. After further adjustment for race, results were similar, but slightly attenuated and household characteristics emerged as the theme with greatest magnitude of association for prevalence of frailty. After additional adjustment for education level results were further attenuated and lost statistical significance (Table 2). After further adjustment for rural-urban residence results were similar for frailty, but for GA impairments results were similar to the race-adjusted model (Table 2).

Table 2.

Prevalence Ratio (PR) and 95% Confidence Interval (95% CI) for the Association between SVI and Frailty and Rate Ratio (RR) and 95% CI for the Association between SVI and Total GA Impairments

		Frailty
		Model 1	Model 2	Model 3	Model 4
Overall SVI	10% Increase	1.08 (1.05, 1.11)	1.06 (1.03, 1.09)	1.03 (0.99, 1.06)	1.04 (1.00, 1.07)
	1 vs. 0	2.12 (1.58, 2.85)	1.77 (1.31, 2.40)	1.32 (0.93, 1.86)	1.41 (0.99, 2.01)
SVI Theme 1	10% Increase	1.07 (1.03, 1.10)	1.05 (1.02, 1.08)	1.02 (0.98, 1.05)	1.02 (0.99, 1.06)
	1 vs. 0	1.88 (1.40, 2.52)	1.67 (1.24, 2.25)	1.18 (0.84, 1.64)	1.26 (0.89, 1.78)
SVI Theme 2	10% Increase	1.06 (1.03, 1.10)	1.06 (1.02, 1.09)	1.03 (1.00, 1.07)	1.04 (1.00, 1.07)
	1 vs. 0	1.84 (1.33, 2.55)	1.72 (1.27, 2.33)	1.37 (0.96, 1.95)	1.43 (1.02, 2.00)
SVI Theme 3	10% Increase	1.06 (1.02, 1.10)	1.04 (1.00, 1.08)	1.02 (0.98, 1.06)	1.02 (0.98, 1.05)
	1 vs. 0	1.82 (1.27, 2.60)	1.49 (1.02, 2.19)	1.20 (0.84, 1.72)	1.18 (0.82, 1.69)
SVI Theme 4	10% Increase	1.04 (1.01, 1.08)	1.04 (1.01, 1.07)	1.03 (0.99, 1.06)	1.03 (1.00, 1.07)
	1 vs. 0	1.55 (1.16, 2.07)	1.45 (1.08, 1.94)	1.29 (0.94, 1.79)	1.38 (0.98, 1.92)
		Total GA Impairments
Overall SVI	10% Increase	1.04 (1.02, 1.06)	1.03 (1.01, 1.06)	1.02 (1.00, 1.05)	1.03 (1.00, 1.05)
	1 vs. 0	1.50 (1.20, 1.87)	1.41 (1.11, 1.79)	1.25 (0.98, 1.60)	1.32 (1.03, 1.69)
SVI Theme 1	10% Increase	1.04 (1.02, 1.07)	1.04 (1.01, 1.06)	1.02 (1.00, 1.05)	1.03 (1.01, 1.05)
	1 vs. 0	1.52 (1.22, 1.90)	1.45 (1.15, 1.83)	1.25 (0.99, 1.58)	1.34 (1.06, 1.70)
SVI Theme 2	10% Increase	1.03 (1.01, 1.06)	1.03 (1.00, 1.05)	1.01 (0.99, 1.04)	1.02 (0.99, 1.04)
	1 vs. 0	1.36 (1.09, 1.71)	1.29 (1.02, 1.63)	1.15 (0.90, 1.46)	1.19 (0.93, 1.52)
SVI Theme 3	10% Increase	1.04 (1.02, 1.07)	1.04 (1.01, 1.06)	1.03 (1.01, 1.06)	1.03 (1.00, 1.06)
	1 vs. 0	1.53 (1.20, 1.95)	1.43 (1.10, 1.87)	1.39 (1.07, 1.81)	1.37 (1.05, 1.79)
SVI Theme 4	10% Increase	1.02 (1.00, 1.04)	1.01 (0.99, 1.04)	1.01 (0.99, 1.03)	1.01 (0.99, 1.04)
	1 vs. 0	1.20 (0.97, 1.48)	1.14 (0.92, 1.42)	1.10 (0.88, 1.37)	1.13 (0.91, 1.41)

Estimated using modified Poisson models with robust variance estimation and generalized estimating equations (for frailty outcome). Estimated using traditional Poisson models with robust variance estimation and generalized estimating equations (for GA impairment outcome). Bold indicates statistical significance at α=0.05.

Model 1: adjusted for age, sex, cancer type, cancer stage

Model 2: further adjusted for race/ethnicity

Model 3: further adjusted for education level

10% increase indicates the model was parameterized according to a 0.1-point increase in SVI

1 vs. 0 indicates the model was parameterized to examine highest (1) versus lowest (0) SVI (indicating highest versus lowest vulnerability)

SVI: social vulnerability index

SVI Theme 1: Socioeconomic Status

SVI Theme 2: Household Characteristics

SVI Theme 3: Racial / Ethnic Minority Status

SVI Theme 4: Housing Type / Transportation

GA: geriatric assessment

Evaluating the models parameterized for highest versus lowest SVI revealed that highest SVI (greatest level of vulnerability) was associated with 2.1-fold higher prevalence of frailty compared to lowest SVI (lowest level of vulnerability) [PR: 2.12, 95% CI: 1.58, 2.85]. Highest versus lowest SVI was also associated with 50% average increase in count of total GA impairments (RR: 1.50, 95% CI: 1.20, 1.87). Again, adjustment for race attenuated these results (Frailty PR [95% CI]: 1.77 [1.31, 2.40]; GA Impairment RR [95% CI]: 1.41 [1.11, 1.79]) and adjustment for education further attenuated the results with loss of statistical significance (Frailty PR [95% CI]: 1.32 [0.93, 1.86]; GA Impairment RR [95% CI]: 1.25 [0.98, 1.60]). Further adjustment for rural-urban status was associated with similar results for frailty, but magnitude of the effect was higher and statistical significance returned for GA impairments (Frailty PR [95% CI]: 1.41 [0.99, 2.01] ; GA Impairment RR [95% CI]: 1.32 [1.03, 1.69]) [Table 2].

There was no evidence of an interaction between education level and SVI (p-value: 0.271). As shown in Supplementary Table 4, there is likely collinearity between race and education status and between SVI and education status. Non-Hispanic White participants were more highly educated than non-Hispanic Black participants and those representing other racial/ethnic identities. Participants with higher education had higher proportions of colorectal and pancreatic cancers, but there was no difference in cancer stage. Finally, participants with higher education had lower proportions of frailty and lower median total GA impairments.

Examining the association between SVI and specific GA impairments indicated that a few specific GA impairments were associated with higher SVI compared to not having the GA impairment. These included: having ≥1 falls in the last 6 months (0.641 vs. 0.554, p-value 0.013), having ≥2 impaired instrumental activities of daily living (IADLs) [0.651 vs. 0.530, p-value: <0.001], having ≥1 impaired activities of daily living (ADLs) [0.652 vs. 0.550, p-value: <0.001], spending most of the day in the bed or chair or being completely bedridden (0.678 vs. 0.558, p-value: 0.010), and having self-reported depressive symptoms most/some/all of the time (0.671 vs. 0.570, p-value: 0.026) [Supplementary Table 5]. These differences constitute an approximate decile increase in SVI based on having these impairments. Supplementary Figure 1 details the proportion of specific GA impairments across SVI quartiles.

DISCUSSION

A 10% increase in SVI is associated with higher prevalence of frailty and higher average count of total GA impairments among older adults with GI cancers prior to systemic therapy. Highest versus lowest SVI was associated with 2.1-fold higher prevalence of frailty and about 50% higher average count of total GA impairments. After adjustment for individual level race, these findings remain with slight attenuation. After further adjusting for education status, statistical significance disappears and associations are attenuated. However, for overall SVI, the association remains at 32% higher prevalence of frailty in the highest versus lowest SVI categories. After further adjustment for rural-urban residence, the findings remained for frailty but were more similar to the second model for GA impairments.

These results suggest that individual-level SDOH, particularly education, may be important in explaining high rates of frailty prior to systemic therapy among older adults with cancer. However, neighborhood-level disadvantage may remain a key predictor. Notably, though, neighborhood of residence is likely a reflection or proxy of an individual’s SDOH.⁴⁰ In fact, prior studies indicate that low individual-level education is independently associated with frailty at baseline and high frailty trajectory over time among a general sample of older adults.⁴¹

This is not to say that neighborhood-level characteristics are not important. In models adjusted for covariates plus the individual-level SDOH of race, SVI is both clinically and statistically significantly associated with higher prevalence of frailty and higher average count of total GA impairments. However, when education is added to the model, statistical significance is lost and clinical significance is attenuated. After examining participant characteristics by education status, education is associated with both race/ethnicity and with SVI. Thus, the attenuation could be related to some collinearity. However, in an analysis of patients with GI malignancies prior to treatment in the CARE registry, NH-Black participants had about 2.6-fold higher odds of frailty compared to NH-White participants even after adjustment for education.²⁰ This, along with continued elevated prevalence after adjustment for education, indicates that despite the attenuation and loss of statistical significance, neighborhood-level SVI may still play an important role in early development of frailty among older adults with cancer. However, the prior analysis did utilize an odds ratio; given high prevalence of frailty in the sample, the odds ratio may over-estimate risk, which is more closely estimated with the PR in the current results. Moreover, the addition of neighborhood-level rurality to the model increased the magnitude of effect further indicating that neighborhood-level characteristics may remain important in early development of frailty. Although there is likely collinearity between rurality, SVI, race, and/or education. Future studies should examine the longitudinal association between SVI and frailty accounting for individual measures of SDOH such as education. Additional individual-level SDOH should also be considered as covariates in future studies. Moreover, causal mechanisms for the relationship between SVI and frailty should be explored to facilitate understanding of key intervention targets particularly given the likely collinearity between many important variables of interest.

Furthermore, prior studies have indicated that neighborhood-level vulnerability (high SVI) was associated with mortality, poor surgical outcomes, and lack of receipt of surgery in the GI cancer setting, all of which may be complicated by presence of frailty.^16,17 However, these studies did not adjust for individual-level SDOH except for race and used county-level SVI as opposed to census tract-level SVI which more closely approximates neighborhood. A published abstract assessed the cross-sectional association between county-level SVI and frailty among older adults with hematological malignancies and found a higher proportion of frailty with increasing SVI quintile (corresponding to greater vulnerability), particularly in the SVI themes of SES and housing type and transportation.⁴² These results are similar to the current findings among older adults with solid tumor malignancies of the GI system utilizing a self-reported rather than an objective frailty index. However, the abstract again utilized county-level SVI, did not include multivariable analyses, and the treatment status of the participants was unclear.

To our knowledge, there have been no studies examining the association between SVI and total number of GA impairments. Our results indicate higher average count of total GA impairments with increasing SVI. However, similar to the frailty results, this association was attenuated and lost statistical significance after the addition of education to the model. Again, collinearity between race, education, and SVI may play a role, but neighborhood and individual-level SDOH may still be working in tandem to produce adverse aging-related effects among older adults with cancer. This is particularly indicated by the addition of rurality to the model, which increased the magnitude of effect and restored statistical significance for SVI overall and themes 1 and 3. Thus, the longitudinal association should again be examined accounting for additional individual-level SDOH and causal mechanisms should be explored.

Should these associations be replicated in longitudinal studies and/or causal studies, interventions for individual-level SDOH or multi-level interventions for neighborhood and individual-level SDOH may be adapted and/or developed to prevent frailty and/or GA impairments among older adults with GI malignancies prior to systemic therapy. Presence of frailty and other GA impairments such as functional decline complicate cancer treatment and may alter clinical decision-making.^6–9 These impairments may also lead to less favorable outcomes and/or premature mortality.^43,44 Interventions to assuage the development of frailty and GA impairments among older adults with cancer are paramount for improving clinical care and outcomes in this population.

This study is not without limitations. First, this is a cross-sectional study so causality cannot be inferred. Specifically, one of the variables used in construction of the SVI is average number of households in the census tract with individuals 65 years or older so SVI and frailty and/or count of total GA impairments could have a bidirectional relationship unable to be explored in this study. While it seems likely that those living in more vulnerable areas would be pre-disposed to frailty or GA impairments, there could be a group of individuals who, because of frailty or GA impairments, are unable to work resulting in lower income and are then forced to move to a more vulnerable area with lower cost of living. Secondly, our sample comes from a single institution from the Southeastern US and results may not be generalizable to other populations. Moreover, our sample is composed of patients accessing care at an academic medical center, which may not be generalizable to the greater socially vulnerable population. Additionally, while we did not find a statistically significant association between cancer stage and SVI quartile on bivariate analysis, it is possible that high SVI could result in later stage at diagnosis which could predispose individuals to frailty or GA impairments. Adjusting for cancer stage in our models likely helps, but limits any inference we can make on the relationship between SVI, cancer stage, and subsequent outcomes. Finally, the outcomes in this study are based on self-report which may induce information bias. However, this is expected to be non-differential across levels of SVI and would therefore bias results toward the null. This study also has several strengths. First, to our knowledge, this is the first study examining the relationship between SVI and frailty and GA impairments among older adults with GI malignancies. Secondly, we were able to adjust for individual SDOH measures in addition to neighborhood-level SVI to account for the multi-level nature of SDOH.

CONCLUSION

Living in a census tract with elevated social vulnerability is associated with increased prevalence of frailty and increased average count of total GA impairments among older adults with GI malignancies prior to systemic treatment. These associations are attenuated with adjustment for individual-level measures of SDOH. Further investigation is needed to determine the key SDOH driving development of frailty and GA impairments and multi-level interventions may be needed to prevent these aging-related outcomes.

Data Availability Statement:

The data underlying this article cannot be shared to protect the privacy of study participants. However, summary-level data may be shared upon request to the authors.